Neuron Explorer

Qwen3-0.6B

line breaks inside running sentences

A neuron in layer 12 of Qwen3-0.6B. Its strongest snippets are very consistent: they mostly peak on the same token. It is featured, so it has steering outputs.

What this neuron does

This neuron fires on a hard line break that cuts a sentence in the middle, such as 'a[newline]newly-minted PhD' or 'reduce the[newline]pressure'. The text from web pages often has these breaks from copy-and-paste. Turning it up did not produce a clear theme.

Written by Claude from this neuron's activations and steering outputs. Confidence: medium. Think it is wrong? Flag it (opens a prefilled issue on GitHub).

Fires most on

The text windows where this neuron's activation was highest. Amber marks positive activation; the outlined token is the peak. Hover a token for its strength.

  1. #1 peak

    ary age, but the greatest concentration is in Cretaceous and Tertiary sediments. Below is a⏎geologic time frame relating to amber formation,⏎based on Jean Langenheim's work

  2. #2 peak

    located in dense tropical jungle and surrounded by valleys and cliffs. In fact, the wind speed rarely goes above 1 m/s. In order to reduce the⏎pressure changes<|endoftext|>Tūtira Restoration Options Presented⏎Scientific modelling of Lake

  3. #3 peak

    is the Tata Employees Union (TEU) formed in 1935. There are two ranks of the workforce at Tata Steel, i.⏎e., officers and non-officers (workers who work in mines and plants and

  4. #4 peak

    on the market products containing microplastic as defined by the REACH restriction.<|endoftext|>« Previous Page⏎Local anesthesia is a method to deliver pain control medicine to⏎a specific area. It can be injected or applied to the skin or mucous

  5. #5 peak

    998. Name: South Korea: Restoration of the Cheonggyecheon River in Downtown Seoul: Executive Summary: Since the 1970<|endoftext|>CS610 Handouts pdf⏎CS610: Computer Networks Hand

  6. #6 peak

    journal entries. It serves as a great way to help sharpen⏎grammar and writing skills.⏎By collecting and recording data on such aspects as water temperature

Steering

Push this neuron up or down while the model writes, and watch what changes. The outputs were generated ahead of time, so nothing runs in your browser.

turned down a lot
Prompt

Here is something I have been thinking about lately: how to make a living wage, and how to define the minimum wage, which could include adjustments for those who are excluded from labor markets, such as entry-level jobs, and future job markets, such as entry-level jobs, there are other

Most negative on

Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.

  1. #1 peak

    this Government instructional activity, students compare and contrast historical Presidential speeches to the ones they hear today. High...⏎9th - 12th Social Studies & History⏎Anticipating the State of the Union, January 2010⏎

  2. #2 peak

    amplification and RNA-Seq.⏎19 Related JoVE Articles!⏎Detection of Rare Genomic Variants from Pooled Sequencing Using SPLINTER⏎Institutions: Washington University School of Medicine, Washington University School of Medicine, Washington University School of

  3. #3 peak

    VE Articles!⏎Assessment of Social Cognition in Non-human Primates Using a Network of Computerized Automated Learning Device (ALDM) Test Systems⏎Institutions: Aix-Marseille University.⏎Fag

  4. #4 peak

    then watch the current year Address and see if their expectations were met. In this State of the Union...⏎7th - 12th Social Studies & History⏎President Obama's State of the Union Address 2011⏎Students pretend

Connections

Show on map

Which neurons in earlier layers feed this one, and which later ones it feeds. Amber links excite, blue links suppress; longer bars are stronger.

Fed by earlier layers

  1. L9:746 weakly excites
  2. L10:157 weakly suppresses end of document and web boilerplate
  3. L1:1857 weakly suppresses
  4. L7:462 weakly suppresses
  5. L10:1747 weakly suppresses
  6. L7:45 weakly suppresses
  7. L8:2420 weakly excites
  8. L9:782 weakly suppresses
  9. L8:1910 weakly excites
  10. L8:967 weakly excites
  11. L11:53 weakly excites
  12. L6:1400 weakly suppresses
  13. L10:65 weakly excites
  14. L11:84 weakly excites
  15. L8:606 weakly excites
  16. L8:625 weakly suppresses
  17. L11:362 weakly suppresses
  18. L3:1192 weakly excites
  19. L1:2012 weakly suppresses
  20. L11:1891 weakly excites
  21. L10:2261 weakly suppresses
  22. L5:965 weakly suppresses
  23. L1:1964 weakly suppresses
  24. L11:2971 weakly excites

Feeds later layers

  1. L27:30 weakly excites
  2. L17:796 weakly excites
  3. L15:434 weakly suppresses
  4. L25:157 weakly excites
  5. L13:3070 weakly excites
  6. L13:1622 weakly excites
  7. L18:2008 weakly excites
  8. L27:1425 weakly excites
  9. L18:1115 weakly suppresses
  10. L20:1538 weakly suppresses
  11. L23:1287 weakly excites
  12. L17:1566 weakly suppresses
  13. L15:1351 weakly excites
  14. L22:964 weakly excites
  15. L18:1275 weakly excites
  16. L14:662 weakly suppresses
  17. L19:2079 weakly excites
  18. L26:633 weakly suppresses
  19. L19:184 weakly suppresses
  20. L15:1810 weakly suppresses
  21. L18:2239 weakly excites
  22. L14:1641 weakly excites
  23. L15:1284 weakly excites
  24. L20:1232 weakly suppresses

Activation histogram

How often this neuron fires at each strength. Almost always it does close to nothing (the tall bars in the middle); the snippets above are the rare moments it fired hard, amber for positive and blue for negative.

1 values between -0.93 and -0.805 values between -0.55 and -0.4340 values between -0.43 and -0.30337 values between -0.30 and -0.181160 values between -0.18 and -0.051719 values between -0.05 and 0.07405 values between 0.07 and 0.2089 values between 0.20 and 0.3239 values between 0.32 and 0.4527 values between 0.45 and 0.5721 values between 0.57 and 0.7014 values between 0.70 and 0.8318 values between 0.83 and 0.957 values between 0.95 and 1.088 values between 1.08 and 1.206 values between 1.20 and 1.331 values between 1.33 and 1.451 values between 1.45 and 1.581 values between 1.58 and 1.701 values between 1.70 and 1.831 values between 1.83 and 1.951 values between 2.20 and 2.331 values between 2.45 and 2.581 values between 2.58 and 2.701 values between 2.83 and 2.961 values between 3.08 and 3.211 values between 3.46 and 3.581 values between 3.96 and 4.08offnegativepositiveoften

Logit lens

The words this neuron nudges the model toward (amber) or away from (blue) when it fires. A hint, not a label: many neurons work through later layers instead.

Promotes

  1. !⏎⏎⏎
  2. auté
  3. !⏎⏎
  4. !⏎⏎⏎⏎
  5. 生
  6. -⏎⏎
  7. !!⏎⏎
  8. !⏎⏎
  9. 年级
  10. 千

Suppresses

  1. icing
  2. rons
  3. yny
  4. frame
  5. BaseContext
  6. отор
  7. mente
  8. RTP
  9. oxy
  10. rogen